Chapter 34

Welfare
 Social Choice

Different economic states will be

preferred by different individuals.
How can individual preferences be
“aggregated” into a social
preference over all possible
economic states?
 Aggregating Preferences

x, y, z denote different economic

states.
3 agents; Bill, Bertha and Bob.
Use simple majority voting to decide
a state?
 Aggregating Preferences

Bill Bertha Bob

More preferred
x y z

y z x

z x y
Less preferred
 Aggregating Preferences
Majority Vote Results
Bill Bertha Bob
x beats y

x y z

y z x

z x y
 Aggregating Preferences
Majority Vote Results
Bill Bertha Bob
x beats y
y beats z
x y z

y z x

z x y
 Aggregating Preferences
Majority Vote Results
Bill Bertha Bob
x beats y
y beats z
x y z z beats x

y z x

z x y
 Aggregating Preferences
Majority Vote Results
Bill Bertha Bob
x beats y No
y beats z socially
x y z z beats x best
alternative!
y z x

z x y
 Aggregating Preferences
Majority Vote Results
Bill Bertha Bob
x beats y No
y beats z socially
x y z z beats x best
alternative!
y z x Majority voting does
not always aggregate
z x y transitive individual
preferences into a
transitive social
preference.
 Aggregating Preferences

Bill Bertha Bob

x(1) y(1) z(1)

y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).

x(1) y(1) z(1)

y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).
x-score = 6
x(1) y(1) z(1)

y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).
x-score = 6
x(1) y(1) z(1) y-score = 6

y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).
x-score = 6
x(1) y(1) z(1) y-score = 6
z-score = 6
y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).
x-score = 6 No
x(1) y(1) z(1) y-score = 6 state is
z-score = 6 selected!
y(2) z(2) x(2)

z(3) x(3) y(3)

 Aggregating Preferences
Rank-order vote results
Bill Bertha Bob (low score wins).
x-score = 6 No
x(1) y(1) z(1) y-score = 6 state is
z-score = 6 selected!
y(2) z(2) x(2) Rank-order voting
is indecisive in this
z(3) x(3) y(3) case.
 Manipulating Preferences

As well, most voting schemes are

manipulable.
I.e. one individual can cast an
“untruthful” vote to improve the
social outcome for himself.
Again consider rank-order voting.
 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1)

y(2) z(2) x(2)

z(3) x(3) y(3)

 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative
y(2) z(2) x(2)

z(3) x(3) y(3)

 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative
y(2) z(2) x(2)

z(3) (3) y(3)

(4) x(4) (4)

 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative and
then lies.
y(2) z(2) x(2)

z(3) (3) y(3)

(4) x(4) (4)

 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative and
(2) then lies.
y(2) z(2) Rank-order vote
results.
z(3) (3) x(3) x-score = 8

(4) x(4) y(4)

 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative and
(2) then lies.
y(2) z(2) Rank-order vote
results.
z(3) (3) x(3) x-score = 8
y-score = 7
(4) x(4) y(4)
 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative and
(2) then lies.
y(2) z(2) Rank-order vote
results.
z(3) (3) x(3) x-score = 8
y-score = 7
(4) x(4) y(4) z-score = 6
 Manipulating Preferences

Bill Bertha Bob These are truthful

preferences.
x(1) y(1) z(1) Bob introduces a
new alternative and
(2) then lies.
y(2) z(2) Rank-order vote
results.
z(3) (3) x(3) x-score = 8 z wins!!
y-score = 7
(4) x(4) y(4) z-score = 6
-score = 9
Desirable Voting Rule Properties
1. If all individuals’ preferences are
complete, reflexive and transitive,
then so should be the social
preference created by the voting
rule.
2. If all individuals rank x before y
then so should the voting rule.
3. Social preference between x and y
should depend on individuals’
preferences between x and y only.
Desirable Voting Rule Properties

Kenneth Arrow’s Impossibility

Theorem: The only voting rule with
all of properties 1, 2 and 3 is
dictatorial.
Desirable Voting Rule Properties

Kenneth Arrow’s Impossibility

Theorem: The only voting rule with
all of properties 1, 2 and 3 is
dictatorial.
Implication is that a nondictatorial
voting rule requires giving up at least
one of properties 1, 2 or 3.
 Social Welfare Functions
1. If all individuals’ preferences are
complete, reflexive and transitive,
then so should be the social
preference created by the voting
rule.
2. If all individuals rank x before y
then so should the voting rule.
3. Social preference between x and y
should depend on individuals’
preferences between x and y only.
 Social Welfare Functions
1. If all individuals’ preferences are
complete, reflexive and transitive,
then so should be the social
preference created by the voting
rule.
2. If all individuals rank x before y
then so should the voting rule.
3. Social preference between x and y
should depend on individuals’
Give up which
preferences betweenone of ythese?
x and only.
 Social Welfare Functions
1. If all individuals’ preferences are
complete, reflexive and transitive,
then so should be the social
preference created by the voting
rule.
2. If all individuals rank x before y
then so should the voting rule.
3. Social preference between x and y
should depend on individuals’
Give up which
preferences betweenone of ythese?
x and only.
 Social Welfare Functions
1. If all individuals’ preferences are
complete, reflexive and transitive,
then so should be the social
preference created by the voting
rule.
2. If all individuals rank x before y
then so should the voting rule.
There is a variety of voting procedures
with both properties 1 and 2.
 Social Welfare Functions
ui(x) is individual i’s utility from
overall allocation x.
 Social Welfare Functions
ui(x) is individual i’s utility from
overall allocation x.
n
Utilitarian: W   ui ( x ).
i 1
 Social Welfare Functions
ui(x) is individual i’s utility from
overall allocation x.
n
Utilitarian: W   ui ( x ).
i 1

Weighted-sum: n
W   ai ui ( x ) with each ai  0.
i 1
 Social Welfare Functions
ui(x) is individual i’s utility from
overall allocation x.
n
Utilitarian: W   ui ( x ).
i 1

Weighted-sum: n
W   ai ui ( x ) with each ai  0.
i 1
Minimax:
W min{ u1 ( x ), , un ( x )}.
 Social Welfare Functions

Suppose social welfare depends only

on individuals’ own allocations,
instead of overall allocations.
I.e. individual utility is ui(xi), rather
than ui(x).
Then social welfare is
W  f ( u1 ( x1 ), , un ( xn ))
where f is an increasing function.
 Social Optima & Efficiency

Any social optimal allocation must be

Pareto optimal.
Why?
 Social Optima & Efficiency

Any social optimal allocation must be

Pareto optimal.
Why?
If not, then somebody’s utility can be
increased without reducing anyone
else’s utility; i.e.
social suboptimality  inefficiency.
 Utility Possibilities
uB
OB

0
0 uA

OA
 Utility Possibilities

uA
uB
OB

0
0  uA
uA

OA
 Utility Possibilities

uA
uB

OB uB

0
0  uA
uA

OA
uB
 Utility Possibilities

uA
uB

OB uB

uB


uA 0
0  uA
uA  uA
uB
OA
uB
 Utility Possibilities

uA
uB

OB uB
uB
uB
uB

uA 0
0  uA
uA  uA
uB
OA
uB
 Utility Possibilities

uA
uB

OB uB
uB
uB
uB

uA 0
0  uA
uA  uA
uB
OA
uB
 Utility Possibilities

uA Utility possibility
uB frontier (upf)

OB uB
uB


uA 0
0  uA
uA
uB
OA
uB
 Utility Possibilities

uA Utility possibility
uB frontier (upf)

OB uB
uB


uA 0
0  uA
uA
uB
OA Utility possibility set
uB
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.

uA
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.

Social
indifference
curves
uA
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.
Higher social welfare

Social
indifference
curves
uA
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.
Higher social welfare

Social
indifference
curves
uA
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.
Social optimum

Social
indifference
curves
uA
 Social Optima & Efficiency
uB Upf is the set of efficient
utility pairs.
Social optimum is efficient.

Social
indifference
curves
uA
 Fair Allocations

Some Pareto efficient allocations are

“unfair”.
E.g. one consumer eats everything is
efficient, but “unfair”.
Can competitive markets guarantee
that a “fair” allocation can be
achieved?
 Fair Allocations

If agent A prefers agent B’s

allocation to his own, then agent A
envies agent B.
An allocation is fair if it is
– Pareto efficient
– envy free (equitable).
 Fair Allocations

Must equal endowments create fair

allocations?
 Fair Allocations

Must equal endowments create fair

allocations?
No. Why not?
 Fair Allocations

3 agents, same endowments.

Agents A and B have the same
preferences. Agent C does not.
Agents B and C trade  agent B
achieves a more preferred bundle.
Therefore agent A must envy agent B
 unfair allocation.
 Fair Allocations

2 agents, same endowments.

Now trade is conducted in
competitive markets.
Must the post-trade allocation be
fair?
 Fair Allocations

2 agents, same endowments.

Now trade is conducted in
competitive markets.
Must the post-trade allocation be
fair?
Yes. Why?
 Fair Allocations

Endowment of each is ( 1 ,  2 ).
Post-trade bundles are
A A B B
( x1 , x2 ) and ( x1 , x2 ).
 Fair Allocations

Endowment of each is ( 1 ,  2 ).
Post-trade bundles are
A A B B
( x1 , x2 ) and ( x1 , x2 ).
A A
p
Then 1 1 x  p2 2  p1 1  p2 2
x

B B
and p1 x1  p2 x2  p1 1  p2 2 .
 Fair Allocations
Suppose agent A envies agent B.
I.e.
( x1B , x2B )  A ( x1A , x2A ).
 Fair Allocations
Suppose agent A envies agent B.
I.e.
( x1B , x2B )  A ( x1A , x2A ).

Then, for agent A,

B B A B
p1 x1  p2 x2  p1 x1  p2 x2
 p1 1  p2 2 .
 Fair Allocations
Suppose agent A envies agent B.
I.e.
( x1B , x2B )  A ( x1A , x2A ).

Then, for agent A,

B B A B
p1 x1  p2 x2  p1 x1  p2 x2
 p1 1  p2 2 .
B B
Contradiction. ( x1 , x2 ) is not
affordable for agent A.
 Fair Allocations
This proves: If every agent’s
endowment is identical, then trading
in competitive markets results in a
fair allocation.
 Fair Allocations
1 OB

2 2

OA 1
Equal endowments.
 Fair Allocations
1 OB
Given prices
Slope p1 and p2.
= -p1/p2
2 2

OA 1
 Fair Allocations
1 OB
Given prices
Slope p1 and p2.
= -p1/p2
2 2

OA 1
 Fair Allocations
1 OB
Given prices
Slope p1 and p2.
= -p1/p2
2 2

OA 1
 Fair Allocations
1 OB

2 2
Post-trade
allocation --
is it fair?
OA 1
 Fair Allocations
1 OB
Swap A’s and
B’s post-trade
allocations
2 2
Post-trade
allocation --
is it fair?
OA 1
 Fair Allocations
1 OB
Swap A’s and
B’s post-trade
allocations
2 2
Post-trade
allocation --
is it fair?
OA 1
A does not envy B’s post-trade allocation.
B does not envy A’s post-trade allocation.
 Fair Allocations
1 OB
Swap A’s and
B’s post-trade
allocations
2 2
Post-trade
allocation --
is it fair?
OA 1
Post-trade allocation is Pareto-efficient and
envy-free; hence it is fair.